Gallery

Here you can view examples of the images I have created in R

Figure 1. Example of machine learning performed in Python. The program identified the plant in the input images (top) and yielded the output images (bottom) to demonstrate which parts of the image have been identified as plant. I could then extract and store data pertaining to the plant. This program was run to obtain data from > 5000 images.

 

 

 

 

Figure 2. Mean growth rate of Garlic Mustard (A.petiolata), colored by genotype and distributed across competition treatments. Using data extracted from images of A.petiolata over a summer (example above), I was able to analyze growth rate in R.

 

 

 

 

Figure 3. Proportion of women reporting cannabis consumption at some point in pregnancy (a), current consumption during pregnancy (b) and intent to consume cannabis while breastfeeding (c). Proportions are distributed across three levels of educational attainment, with and without partners who consume cannabis. For each group of women, coloured bars represent the proportion of women who reported the outcome and lines represent the 95% confidence intervals around each proportion.

 

 

 

 

Figure 4. Fitness reaction norms for the relationship between enemy abundance (E) and fitness (W) for a susceptible (light orange) and a resistant (blue) genotype. The three columns show different scenarios in which the defence cost—benefit ratios are high (a,d,g), moderate (b,e,h) or low (c,f,i). The three rows show different scenarios for variation in the abundance of enemies in the native (N, green) and introduced ranges (I, yellow). Of these nine scenarios, only one supports the evolution of increased competitive ability (EICA) in which natural selection for defence reverses between ranges (panel e).

 

 

 

 

Figure 5. Association of leaf quality (PC1) with Alliaria petiolata rosette size and fecundity. Green circles denote A. petiolata plants grown in the alone treatment, orange squares denote A. petiolata plants grown in the interspecific treatment and blue triangles denote A. petiolata plants grown in the intraspecific treatment. The colored lines show the estimated effect of leaf quality on rosette size and fecundity (i.e., selection gradients), with color corresponding to treatment. In the case where the effect was not significant, a slope of zero is shown to reflect the null hypothesis that the slope is equal to zero.

 

 

 

 

Figure 6. A general defence trade-off model of defence traits along a gradient of enemy abundance. Fitness reaction norms (top row) show the relationship between enemy abundance (E) and fitness (W) for five different genotypes with defence allocation ranging from low (light red) to moderate (green) to high (dark blue). The three columns show different scenarios in which the defence cost—benefit ratios are high (a,d), intermediate (b,e) or low (c,f). The bottom row shows the variance in fitness for each of the three scenarios, with high variance corresponding to a rapid evolutionary response to selection.

 

 

 

 

Figure 7. Simulated data demonstrating the predicted relative fitness (W) when enemies are present (E+) and absent (E-) across a gradient of susceptible (red) and resistant (blue) genotypes.

 

 

 

 

Figure 8. Linear approximations of reaction norms for fitness (W) measured in a susceptible (S, red) and a resistant genotype (R, blue). The difference in y-intercepts (WS0 – WR0) provides an estimate of the cost of resistance, while the slopes (–W0/E0) estimate of the benefits of defence.

 

 

 

 

Figure 9. Effect of competition on A.petiolata and A.saccharum under different competition regimes.